NEWS
%?S*
Almost four years after the Workshop
on End-S tage Transition (Curr. Sci.,
1994, 67, 6-9; Proceedings of the
Workshop) now seen as Minnowbrook
I, nearly 50 participants gathered again
at the same lovely site on the initiative
of the same organizers (John Lagraffe of
Syracuse, Paul Gostelow of Leicester
and Terry Jones of Oxford). As Lagraffe
reminded his audience in his opening
remarks, the first workshop had sought
to bring together two different communities involved in transition research:
the J o w n a l of Fluid Mechanics and
American Society of Mechanical Engineers types, or high and low church, as
Paul Gostelow had described them. The
present workshop had a sharper Focus
on turbomachinery flows.
Roddam Narasimha (JNC/IISc, Bangalore) started the proceedings by reviewing what had transpired at and after
Minnowbrook I. The distinguishing
features of transition in turbomachinery
flows are that they are unsteady on time
scales longer than typical eddy turnover times, occur in harsh and highly
disturbed environments at awkward
Reynolds numbers, and are generally
three-dimensional (even in the mean).
As a consequence, the flow is a veritable -fluid-dynamical ‘zoo’, characterized
by separation, reattachment, transition,
relaminarizati o n , ret r ansi t ion, etc., all
often occurring in the same flow. Greg
Walker (Tasmania), in his keynote talk
the next day, noted how ‘interesting’ the
problem was, and that this may be a
version of the well-known Chinese
curse about living in interesting times
for turbo -ma c11 i n e r y fl u i d dyn ami ci st s ,
Nevertheless, as Narasiinha noted, there
had been inuch progress on many fronts,
but it was remarkable that ideas related
to dynamical chaos, which at first appear so closely connected with transition to turbulent flow, had made so little
impact on the subject they were going to
discuss at the meeting: the reduction in
the number of degrees of freedom that
such ideas have till now provided is
clearly not sufficiently significant to
make a practical difference.
”Report on a Workshop on Boundary Layer
Transition i n Turbomachines held from 7 to
10 September 1997 at Minnowbrook, New
York, USA.
274
Minnowbrook I had been dominated
by discussions on spots and modelling.
Minnowbrook I1 saw much attention
devoted to the characteristics of the
calmed zone trailing a spot and the
possibilities of its exploitation, and the
understanding and management of separation bubbles and unsteady transition.
But how important is transition research for the turbomachinery industry?
Walker cited GE compressor tests (made
by Halstead) showing transition extending over GO% of the blade chord, and
estimates of potential improvement in
efficiency by several percentage points;
consider i n g h ow w i d ely tu r b o machi n es
are used in energy conversion and propulsion systems, significant economic
and environmental benefits are possible.
David Wisler (GE Aircraft Engines,
Cincinnati) pointed out that the ‘lack of
ability to predict the location of boundary layer transition for components in
gas turbine engines is impeding our
ability to gain maximum benefit from
our design effort’. If a complete computational fluid dynamics (CFD) design
tool incorporating transition were to be
available, h e foresaw airfoil designs
with higher blade loading that would
reduce part count and improve efficiency. He estimated that a 1% improvement in the efficiency of a low
pressure turbine would result in a saving
of $52,000 per year on a typical airliner,
but he was willing to accept even 1/2%
provided it was reliable. Improved
transition technology was thus very
relevant. However, Wisler thought it
was unlikely to have a major impact on
design: the gas turbine industry is now
maturing, and is generally driven by
cost and ‘error-proofing’ the design
rather than higher performance per se.
Apart from the gains that may be
achieved if a reliable CFD design tool
were to be available, where should one
look to improve efficiencies now? The
answer, as summarized by Greg Walker
and widely endorsed during the meeting, appears to lie in the exploitation
of the calmed region trailing a spot,
understanding the time-dependent forcing of transition in a turbine row because of wakes from upstream rotor
stages, and the management of separation bubbles.
cons id ered
N aras i m h a
threedimensionality important, and David
Ashpis (NASA Lewis) listed 3D effects
(‘virtually unknown’ in his words) as
one of the barriers to progress. Several
part i ci p an t s high 1igh t ed transit i on , sep aration and wake-boundary layer intcraction as the tnajor problems in
t ur boinachinery boundary layers. As hpi s
noted a 2% drop in efficiency From takcoff to cruise (as the Reynolds number
drops typically from about 200,000 to
80,000), and minimizing this loss was
one of the objectives of the NASA Low
Pressure Turbine (LPT) Flow Physics
Program (which involves 6 industries, 6
universities and 3 government laboratories).
The calmed region
The possibility of exploiting thc calined
region had been foreseen already in the
pioneering work of Schubauer and Klebanoff, who wrote in 1955 [as Howard
Hodson (Cambridge) reminded us] that
‘turbulence injected at t h e proper lime
intervals can in principle alleviate the
severity of the turbulence “disease”’. As
of today there is no clear demonstration
of the efficacy of this ‘vaccination’
principle - as we may call i t - but there
are now indications of how it might be
done.
Terry Jones and John Lagraffe described work showing that the behaviour of the calmed region could be seen
as the resumed growth of a disturbed
laminar layer; the rate at which the
calmed region trailing c? spot grows is
defined by the time for a new boundary
layer to grow from the point where t h e
spot was born. This view leads to a
good prediction of the duration of
calming, and a collapse of measured
heat flux data as a function of time normalized by the duration so estimated.
Paul Gostelow pointed out that as the
flow in the calmed region is more stable
than the boundary layer it replaces, i t
delays separation and suppresses instability. In experiments on a flat plate
subjected to strong adverse pressure
gradients, Gostelow finds that the calm
region behind a triggered spot i s extensive. Its interaction with the natural
CURRENT SCIENCE, VOL. 74, NO. 4,25 FEBRUARY 1998
NEWS
boundary layer is complex, and depends
on whether the boundary layer is laminar o r turbulent. He introduces a new
relaxation parameter to quantify and
describe the calmed region, and uses the
data obtained to validate his model for
transition.
Hodson showed some experimental
data indicating that a higher frequency
of inducing transition spots could reduce the loss coefficient and trailingedge momentum thickness. Ari Seifert
(Tel Aviv) reported on experiments
showing that a train of spots generated
some 200 momentum thicknesses zipstream of the transition onset location
can lengthen the transition zone and so
significantly reduce the intermittency
downstream in the transition zone;
whether this would compensate for the
iizci-ease i n intermittency upstream and
so actually reduce drag is not yet clear.
Unsteady transition
The time-dependent forcing of transition
already referred to attracted considerable attention at the meeting. Pave1
Jonas (Prague) reported interesting
measurements on the transilion zone at
different turbulence levels with superposed periodic oscillations of freestream velocity (obtained by rotating a
flap at the trailing edge of his testsurface). When the frequency of oscillation is low, tlie final part of the transition zone is considerably longer, and
the (average) skin friction shows nonmonotonic variations in space; but i f the
frequency is high the transition is rapid
as in a turbulent-free stream. The
changes appear l o correlate with an unsteady Reynolds number directly proportional to the velocity amplitude and
inversely to the frequency, with a critical value that determines whether the
frequency is ‘high’ or ‘low’. Jonas also
reported on the effect of turbulence
length scale, which he could vary over a
range of 1:15 in his facility without
altering the turbulence intensity; he
found a delay i n the later stages of
transition when the scales were small.
Hodson reported experiments using
surface hot films from unsteady cascade
as well as full scale rig testing in an
altitude test facility. David Halstead
(GE Aircraft Engines, Cincinnati)
showed detailed measurements of flow
field unsteadiness i n multi-stage com-
pressors and turbines. These experiments provide not only the intensity and
the spectrum, but also measurements of
the length scales in a low pressure turbine. The complexity of the unsteady
flow field between blade rows was
shown to increase markedly through the
turbine. At the inlet of the second stage
nozzle, the unsteadiness correhtes with
the clocking position of the upstream
nozzle blade row, (Such correlations
might ‘select’ particular blades for failure, as Ashpis noted.) At the exit of the
turbine, however, no such correlation
could be found, It is clear that the wakes
from stator and rotor blades upstream
get chopped up, and that at some stage
in the machine identification of the origin of various chunks of turbulence will
become difficult, but it is not clear how
deep into the inachine one has to go
before this happens. At the exit of the
turbine there is no discernible ‘freestreain’ region, suggesting that experiments with moving wake generators
upstream of a cascade or single stage
will not simulate many features of the
ilow field in a multi-stage machine.
Om Sharma (Pratt and Whitney, E.
Hartford) showed experimental data
from an engine and a inodel rig to highlight current lack of understanding of
the actual operating environment in the
gas turbine engine. Analytical results
showed the limitations of transitionlturbulence models now in use in
the low Reynolds number environment
of low pressure turbines; this problem is
important because of the loss in efficiency during cruise at altitude, where
Reynolds numbers tend to be low relative to takeoff conditions.
Separation bubbles
Separation bubbles are another characteristic Feature of many turboinachine
blade flows. Ting Wang (Clemson)
sketched three hypothetical modes of
transition involving bubbles: (i) the
short bubble at low Reynolds numbers
(Re) and mild adverse pressure gradients, (ii) the long bubble - also at low
Re but in strong adverse pressure gradients (in both cases the separating flow is
laminar), and (iii) the transitional bubble. In the short laminar bubble, transition onset accurs
at maximum
displacement point, and maximum turbulence intensity (u’) at (turbulent) reat-
CURRENT SCIENCE, VOL. 74, NO. 4,25 FEBRUARY 1998
tachment. In tlie long bubble, there is a
short, early transition region leading to
the first reattachment type behaviour,
but the strong pressure gradient does
not admit attached flow; a long and late
transition occurs in the bubble before
actual reattachment. The transitional
bubble occurs at high R e and mild adverse pressure gradients. Here transition
begins before separation and tends to
house a long initial region (because of
the constant pressure plateau induced by
the bubble), and is followed by a shorl
completion phase. Maximum it’ occurs
near the maximum displacement point.
Ting Wang presented correlations for
the Re values associated with bubble
events as a function of an acceleration
parameter for all three modes.
Nick Curnpsty (Cambridge) presented
measurements showing how lree stream
turbulence and incidence have a large
el‘fect on the naturc of the separation
bubble. At low incidence no evidence of
spot transition could be seen, but at
higher incidence and low free stream
turbulence, spots were clearly visible in
the bubble shear layer.
Fred Simon expressed surprise that
even in separation bubbles the intermittency followed Narasiinha’s universal
distribution, although the Emmons spot
was presumably specific to wallbounded flows. Narasimha explained
that the universal distribution depended
on three postulates: Poisson birth, linear
propagation and concentrated breakdown; and if these were satisfied the
universal distribution follows, wliether
or not the Emmons spot is involved,
Thus, he had found that the temporal
development of phase transition in
metals followed the same or related
distributions.
Stability
Eli Reshotko (Case Western) surveyed
the role that transient growth (first invoked by Landahl in 1980 to explain the
lift-up mechanism in boundary layers)
can play, especially in roughnessinduced transition (where there may be
no evidence of TS waves), and presented a new picture of possible paths to
turbulence in wall layers. In the canonical TS route such transient growth is
benign and insignificant. Other routes
are that transient growth provides
higher input to eigen-mode growth or
275
NEWS
directly excites secondary instabilities.
Finally, transient growth can feed into the
bypass mechanism. With very large
forcing possessing a crazy spectrum,
Reshotko thought bypass mechanisms
can be directly excited without transient or
eigen-mode growth. Interestingly he found
evidence of transient growth in pipe experiments conducted at JPL in 196 1 !
Herbert reported that the PSE codes
that work so well for external flows are
less successful for turbomachinery
blades, Among various possible causes,
he found there could be a serious problem even in the specification of the geometry of the blade (usually coordinates are given at too many points
but to too Sew decimals). Further, computed pressure distributions do not
agree with measurement (blockage corrections? gradients in outer inviscid
flow? non-zero wall-normal pressure
gradients?); these problems need to be
sorted out first. At any rate, his codes
have now seen various improvements
like extension to swept-wing flows and
inclusion of Klebanoff modes.
Narasimha briefly described work
done with R a m Gavindarajan (NAL,
Bangalore) on a new formulation of
boundary layer stability. This work
leads to a hierarchy of three equations,
most easily formulated for the FalknerSkan similarity solutions. The highest
member of the hierarchy is compatible
with Herbert's PSE, but at this order
consistency in approximation demands
that account be taken of higher-order
boundary layer theory, which presents
difficulties and is therefore generally
(but unjustifiably) ignored, The next
member of the hierarchy is a lower order parabolic (LOP) p.d.e. that is simpler than the Herbert PSE; this is the
highest order rational theory that can
justifiably neglect higher order boundary layer effects. At sufficiently large
Reynolds numbers, this LOP theory
reduces to an 0.d.e that is like OrrSornmerfeld but not identical: two viscous 0-S terms are absent but an additional term representing the transport of
vorticity by the mean wall-normal velocity is present. All three members of
the hierarchy apply to non-parallel flow.
Results show significant differences
from Orr-Sommerfeld only in strong
adverse pressure gradients; in free shear
flows the boundary of absolute instability can be substantially different, and
276
this may be important for understanding
separation bubbles.
Frank Smith (London) reported theoretical studies on near-wake transition,
inspired by the need to understand rotary-blade/wake interactions. The nearwake may exhibit two length scales,
bechuse of the presence of a thin inner
sublayer within a thicker boundary
layer. He noted that a favourable pressure gradient may destabilize a near
wake because of the negative profile
curvature it induces. Smith higlilighted
various issues in which the stability
characteristics of the wake - absolute or
convective, linear or nonlinear - may be
relevant to transition problems on
blades.
Modelling
Models for the transition zone are now
proliferating rapidly, and are being generalized to the unsteady situations prevailing in turbomachinery. Narasimha
described briefly an improvement made
by Dey (IISc, Bangalore) to linearcombination models, so that they can
take account of the calming period and
ensure better momentum conservation.
Fred Simon (NASA Lewis) presented
models that utilize the dependence of
spot characteristics on pressure gradient
and Mach number for predictions of
heat transfer; he uses the intermittency
model of Solomon, Walker and Gostelow and correlations for the nondimensional spot formation rate of
Narasimha. He finds that relative
changes can be estimated well independently of the model, but it is important to incorporate Mach number
effects. His model gives good agreement
with the data of Arts et al.
Johnson (Liverpool) uses a distributed breakdown model, postulating that
spots are induced whenever the instantaneous velocity drops 50% below the
mean - inducing transient local separation (an idea that I must point out goes
back to G.I. Taylor). By making suitable
assumptions on the proportion of minima that induce spots, Johnson provides
a good fit to the intermittency near onset. With correlations for the spot format'ion rate parameter of Narasimha, he
is able to get good agreement with the
data of Abu-Ghannam and Shaw and
Gos telo w.
Fraser (Dundee) has implemented a
transition model within the PHOENICS
package. This is a linear-combination
integral model, with correlations for the
start of transition. The integral mode1 is
a set of coupled ordinary differential
equations, generated from a KE type
equation making certain simplifying
assumptions. Comparisons were shown
for several ERCOFTAC test cases, but I
noted that the peak skin friction coefficient is always under-predicted.
Lakshminarayana (Penn State) described recent work on a low Reynolds
number KE model; he found the FanLakshminarayana-Barnett version gave
the best results. The model is able to
predict the occurrence of the major regions associated with wake-induced
transition, including the calmed zone
and the transition strip.
Steelant (Gent) has pursued the idea
of modelling by-pass transition Llirough
conditionally averaged flow equations,
each carrying interaction terms that
depend on the conditionally averaged
velocities as well as the intermittency y,
which is determined either algebraicaIly
(after Dhawan and Narasimha) or
through a separate transport equation.
Reasonable agreement was reported
with the ERCOFTAC flow T3C5.
Guohua Xiong (Kentucky) also uses
conditioned KE equations with a transport equation for y , to predict transitional flows in low-pressure turbines,
When an allowance is made for the
normal variation of y through the Klebanoff distribution, Xiong finds good
agreement with the ERCOFTAC turbine
series of test cases.
Daniel Dorney (GMI Institute, Flint
MI), using Baldwin-Lomax, KE and qw
models, reported on both cascade and
stage simulations of unsteady transition.
The first two models give similar results
on cascades. At a Reynolds number of
80,000, a transition limit cycle appeared
in the solutions. They remind me of
Rotta's and Pantalu's observations in a
pipe, but it is not clear if there is a connection.
Mark Savill (Cambridge, not present
at the meeting) circulated a useful report
about the coordinated European programme being promoted by the
ERCOFTAC
Transition
Modelling
Group. The programme now includes 15
groups outside Europe as well, and is
becoming a major international force for
CURRENT SCIENCE, VOL. 74, NO. 4, 25 FEBRUARY 1998
NEWS
assessment of transition models - often
carried out ‘blind’, i.e. without access to
test data, or even in advance of its acquisition. A total of 15 test cases have
so far been released to participants.
Savill’s report for Minnowbrook contains a detailed description of model
performance in various cases.
DNS/LES
In his opening remarks, Narasimhn
made what he called a ‘modest proposal’, which was to do Direct Numerical Simulation (DNS) (and possibly also
Large Eddy Simulation (LES)) on a
realistic blade, but in the manner advocated in Minnowbrook I (i.e. in ‘mission
mode’). He suggested beginning with
the 2D
case, with
progressive
‘complexi fication’ in the geometry (to
3D), specification of the disturbance
environment (wake-passing), inclusion
of compressibility effects, cooling, etc.
The current state of the art was discussed in several presentations. Manmohan Rai (NASA Ames) simulates the
spatially-growing boundary layer on a
heated flat plate i n the presence of relatively high free-stream turbulence. The
simulation involves 24 x lo6 grid
points. Thc basic physics of transition is
captured, but computed onset is slightly
downstream of observed onset in the
experiments of Sohn et al. (which Rai
simulates). It is not clear why the fully
turbulent boundary layer far downstream appears to originate at the leading edge, and has a profile without a
wake region. Interestingly, Rai found no
evidence of negative eddy thermal f u x ,
of the kind that some experiments have
(controversially) reported.
Neil Sandham (Queen Mary and
Westfild) has now carried out simulations of separation bubbles showing
laminar separation, transition and turbulent reattachment, on a 128-processor
Cray T3D (achieving 90% parallelization efficiency and needing 20,000
processor-element hours at around 3
gigaflops). Sandham considered that
fully-resolved DNS of separation bubbles, and single and multiple spot calculations, are now feasible.
D. J. Doorly (Imperial College) uses
the vortex particle-in-cell method to
study wall-layer dynamics and vortexsurface interactions. In a model problem, a 2D vortex suddenly imposed on a
wall is shown to lead to eruption of
boundary-layer vorticity into the free
stream.
Miscellaneous
William Saric (Arizona State, Tempe)
summarized the detailed experiments
being carried out in the ASU Unsteady
Wind Tunnel to understand receptivity
to free stream disturbances, in particular
tlie large-amplitude noise characteristic
of turbine engines; the problem cannot
be handled through the usual linear
mechanisms. 0bli qu e and broad- band
sound waves are used to determine how
unstable waves are initiated. Acoustic
disturbances are produced in bursts, and
the TS waves detected by a differential
phase speed technique. Saric emphasized that theory, DNS and experiment
are now in agreement, provided due care
is exercised in defining receptivity coefficients.
As Walker pointed out, transition is
most important in the LP turbine and
compressor, where the blades have relatively high aspect ratio; low aspect ratio
blades are largely immersed in the turbul en t annulus wall boundary 1 ayer.
This - and experimental convenience have led to a large effort in 2D flows.
But true 2D flows are rare, and 3D phenomena could be very significant.
Narasimha showed results of measurements (made with Jahanmiri and
Prabhu) on a spot i n a divcrgent flow
with no pressure gradient; i t was found
that the spot can cut across local
streamlines at angles ranging from 3” to
13”. On the other hand the spot wedge
grows at an included angle of 20°, about
the same as in 2D flow. He concluded
that divergence only produced a geometric distortion of the coherent structure in the spot (it was no longer
symmetrical), but hardly affected its
dynamics. Narasimha also mentioned
the investigation of the 3D transition
zone in the same diverging flow carried
out by Ramesh, Dey and Prabliu (IISc).
Kohama (Sendai) reported detailed
investigations of transition on a swept
flat plate with a displacement body. He
found two different instabilities apart
from the well-known stationary mode:
namely a travelling cross-flow instability and a high-frequency secondary instability
propagating
along
thc
stationary vortex. The turbulent wedge
CURRENT SCIENCE, VOL. 74, NO, 4,25 FEBRUARY 1998
triggered by this instability starts from
the middle of tlie boundary layer on
each cross-flow vortex. Koharna also
reported on experiments at control using
suction.
Tumin (with Eliahou, Han and Wygnanski at Tel Aviv) showed that transition in pipes begins with the distortion
of the parabolic velocity profile, and
proceeds through stream-wise rolls
which amplify and finally break down
following spikes as in the K-regime in
boundary layers.
Jacques Lewalle (Syracuse) showed
how wavelet analysis can be used on hot
film series data to characterize and track
structures in the flow.
McEligot described a new matchedindex-or-refraction (MIR) facility that
has been built at the Idaho National
Engineering and Environmental Laboratory. The idea is that if fluid and model
in a flow facility have the same refractive index, one can use laser Doppler
and particle imaging velocimetry techniques to see through objects without
disturbing the [low. Problems connected
with transition on cascades and flow in
cooling holes are proposed to be investigated.
Recommendations
As at Minnowbrook I, working groups
were formed to look into selected areas
and recommend directions for future
research. Terry Jones, who headed the
Working Group on the Becalmed Region, pointed out that details at the
trailing edge of the spot are not yet understood; e.g., why does it travel at half
the free-stream speed? More work needs
to be done on practical exploitation, not
only to help manage separation, but
possibly also aeroelasticity and RCOUStics.
David Ashpis, whose Working Group
considered LPT Physics, called for an
‘Engine Flow Environment Measurement Project’. The discussions at Minnowbrook clearly brought out the
serious need for such a project. Ashpis
further suggested that the project should
be international - in execution as well
as in the review process. He also highlighted the need for further studies of
Mach number and (severe) adverse pressure gradient effects on spots.
Terry Simon reported on behalf of
end-users. What industry needs is robust
27 7
NEWS
transition/turbulence models for offdesign analysis to evaluate design
choices and determine optiinum designs.
(‘Robustness’ was defined by Wisler as
the ability to work over a wide envelope
of initial and boundary conditions,
Mach and Reynolds numbers, and geometry, without needing special tuning.)
Such models have to be integrated into
the standard CFD codes. But, interestingly, the Group added that it is necessary to include a presentation of
the physics of transition as well, to
help engineers in design selection. Finally, the importance of desensitizing
the engine to low-Re effects was emph asized.
The fourth Working Group on computations was chaired by Sandharn, The
Group identified the following DNS
tasks as feasible in 3 years on current
computer platforms:
-
-
Spots, in a range of pressure
gradients, with study of sensitivity of results to method of triggering.
Separation bubbIes on a single
blade, at Re = 50,000, M = 0.8. Om
Sarma offered cascade B as a candidate, noting that further experimen ts with we1I-defin ed boundary
conditions may need to be carried
out.
In the 3-10 year time frame, assuming
computer power increases 10-fold in 3
years, it might be possible to analyse
wake-blade interactions. Doorly felt
this would still be expensive. Herbert
considered that the biggest problem
would be to prescribe the appropriate
inflow boundary conditions at the HP
turbine.
There was much discussion on how
all of this work was to be funded.
Ashpis suggested that NASA, Department of Energy and the Air Force
office of Scientific Research need to
formulate a joint strategy For supporting long-term programmes. This is all
the more necessary as there are now
fewer people i n industry working
in ‘enabling technology groups’, so
the task has to be accomplished by academia, government and industry working together for visible benefits.
Industry was willing to talk, Terry
Simon said, if academics were willing to
listen !
27 8
Conclusion
In his concluding remarks, Narasimha
said i t is even clearer today than it was
in 1993 that turbomachinery boundary
layers represent a very interesting and
complex flow situation. There are other
flows in the same awkward Reynolds
number range, such as birds and
windmills, but these do not present the
complexity that arises from the multiple
stages in turbomachinery. The industry
is now mature, and costs seem to be
driving everything: it is willing to consider even a half per cent improvement
in efficiency, but will not easily pay for
attaining it themselves.
As far as transition physics is concerned, the increasing interest in separation bubbles is both timely and
welcome. With the many interesting
studies reported at the meeting, it was
his assessment that in the corning few
years the bubble problem would be
largely sorted out, except of course for
all .those more basic questions that
might remain in connection with spots
and turbulence.
With regard to spots it was gratifying
that the studies that he had foreseen in
1985 as essential were now being conducted. The behaviour of spots in pressure gradients, and the effect of Mach
number, are slowly getting to be documented. But it is important to realize
that we still d o not understand spots.
There is nothing more practical than a
good theory, as Bbltzmann had said, but
one is not in sight yet: and Frank Smith
is the only brave soul tackling this difficuIt problem.
The other interesting development
over the last four years has been the
considerable attention being devoted to
‘the calm that follows the storm’. The
possibilities of exploitation of the calm
region trailing the spot, foreseen long
ago by Schubauer and Klebanoff, are
now being more quantitatively assessed.
The simpler situations in unsteady
periodically-forced transitions are being
sorted out. There is a big question regarding the disturbance environment in
turbomachinery, and the proposed project on the subject was absolutely essential in his view. At the same time, as one
moves deeper into the turbine and upstream wakes get chopped up, it is unlikely to be either necessary or desirable
to follow each wake all the way
through. Surely there must at some stage
be a method of characterizing the disturbance environment statistically once
again, although the statistics required
would not be just the intensity, or even
a length scale or a spectrum, but would
have to be event-based: are there lumps
of turbulence? at what rate do they arrive, how fat and intense are they? - and
so on.
Transition models are now proliferating. Considering that even five to ten
years ago there were only a few models
around, growth has been rapid, and the
subject is now becoming a minor industry. A major question that still remains
is whether the models explicitly include
the intermittency or not, and if they do
whether it is treated as a parameter or as
a dynarnical variable. Models now exist
adopting each of these options.
Narasimha confessed to a personal preference for physical, integral methods,
where one sees exactly what is being
done, in contrast to the more elaborate
multiple p.d.c. systems, the reason being
that ideas for control, management,
manipulation, etc. are more obvious
with the integral models. Of course the
p.d.e. systems will perhaps integrate
more easily with what is now being
generally done with fully turbulent
boundary layers, and if that route is
followed, it would b e essential in his
view to follow the suggestion of David
Ashpis and provide the designer with a
presentation of the physics as well, so
that he might be able to pick more intelligently the design options that
should be explored. Whatever madel is
adopted, it is d e a r now that i t would
have to include unsteady transition as
well.
Narasimha was of the firm view that it
was only a matter of time before DNS
would start to be used in a practical
way, even though it was not going to be
easy and may not be immediately
achievable. The main reason f a r this
view was that the Reynolds ntimbers in
turbomachinery are relatively low: even
Re = 50,000 is of direct practical interest. He predicted that the first major
field in technology where DNS wiIl
make a direct impact will be turbomachinery. It was therefore prudent to
prepare for the future, and if possible to
make it happen: after all, computer
power will probably have increased
about a thousand-fold or more in 10
CURRENT SCIENCE, VOL. 74, NO. 4,25 FEBRUARY 1998
NEWS
~
years' time. He did not want to suggest
that the full turbine or compressor
should b e handled by DNS, but there
were various bits of the system that in
his view should be, keeping always the
designer's needs in mind.
He referred to questions often raised
about the possible relevance of stability
theory in turbomachinery. At first sight
stability would seem unlikely to be a
major consideration because of the high
disturbance environment, but various
presentations made at the meeting
showed that such a conclusion might be
hasty. Reshotko's work on algebraic
growth showed the importance of analysing transient disturbances. There was
still the unresolved question about why
en methods do not work as well on
blades as on wings; Herbert had suggested several reasons that need to be
investigated. Narasimha saw further use
for the simpler alternatives that he and
his colleagues at Bangalore had investigated. Questions of global and convective instability may play an important
role in achieving a fuller understanding
of the behaviour of separation bubbles
and of near wakes, as Frank Smith had
suggested.
Narasimha finajly listed what appeared to him to be conspicuous gaps in
the present research scene. There is first
of all too little work on threedimensional flows, too little DNS/LES
(done specifically for turbomachinery
flows), and too little theory, A coorclinated project on a given blade row as
more and more stages are mounted upstream - coordinating between experiments and computation, if not theory seemed to him to be a great need. Along
the line of thc suggestions made by
Ashpis, international programmes on the
CURRENT SCIENCE, VOL. 74,NO. 4,25 FXBRUARY
disturbance environment in turbomachinery and on DNSILES would seem
highly worthwhile; ERCOFTAC and
Minnowbrook have shown how such
international programmes can be organized. With the funding situation becoming so difficult, there seemed to be no
alternative.
He concluded by thanking the organizers for putting together again such an
interesting and pleasant meeting where
industry, academics and government
were all so well represented and interacted with such enthusiasm and frankness.
Roddam Narasimha, Fluid Dynamics
Unit, Jawaharlal Nehru Centrc for Advanced Scientific Research, Bangalore
560 064, India and Centre for Atmospheric
and Oceanic Sciences, Indian Institute of
Science, Bangalore 560 012, India.
27 9